CD4+ T-help Cell Fate Optimization Using the D-WaveSys Quantum Processing Unit

This thesis introduces quantum computing. Rather than an abstract view of it, which may be somewhat overwhelming to a neophyte (as I still am because of the tremendous information that’s available), this work contains an example application that may hopefully be useful to those in the field of healthcare and  immunology. There are extremely few U.S. universities

that offer even a single class in Quantum Computing per se, much less an advanced degree in it. Notwithstanding “Quantum Computing”, Southern California only has two schools that offer degrees in Quantum Information Science, and those are USC and UCLA. (I’m confident that this school, CSUN, will soon offer classes in quantum computing.) The virtue of quantum computation is evident in a vast spectrum of real-world problems ranging from aerospace to zoology, (“a” to “z”). The practical example I present, is the plasticity (functional change – see pg 1) of a specific type of human white blood cell. It’s hopeful to make it useful for the common-good. User input is required. Details as to the process is evident in this paper, as well as in the Python and quantum coding.

The learning curve is intentionally very gentle in this work, and incrementally builds on previous knowledge bases to eventually explain the applied quantum project I present here. Careful consideration is given to the biological portion of this thesis. Since this is not research in biology, but rather a computer science endeavor, this work illustrates and applies a recent research groups’[2] immunological concept. I bring it into the quantum computing domain. The reason for employing a QPU (quantum computing unit processor), is because of the rather intricate, nuanced,  and complex relationships between many proteins that are related to this single specific white blood cell just mentioned [pg vii] ie., the human Thymus cell type, “CD4+ T-help”. Please recall that abbreviations are listed starting on page v. 

https://github.com/TomEphraimPerez/learnleap

1. Introduction

We contain many different types of lymphocytes, (WBCs or white-blood-cells, see Fig-1).Fig-1 presents a hierarchy of some cell types in this work, and is a quick and useful reference.

As an illustration when we contract a flu, we may experience discomfort from our swollen lymph nodes where lymphocytes living there and are called thymus cells, or T-cells. One important T-cell is called CD4+ T-help.  CD4+ T-cell is a very important lymphocyte that we can’t live without. ”CD” means cluster of differentiation [18][19]. T-help cells (Th), Macrophages, and Dendritic Cells [20][21]) when signaled, make Antigen Presenting Cells killing harmful antigens. (The lack of mention of “pathogen” is intentional, and out of scope for this work.) The CD4+ T-cell is the focus of this paper but moreover, quantum computing is the centerpoint. This research is two-fold.

Firstly:

This work demonstrates the necessity and applicability of quantum computing. The focus of Carbo et al [2] and inspiration for this work, is the CD4+ T-cell. This includes a possible heterogeneous subset (Fig-2). A mathematical  optimization can be obtained for the subset. This will be done by a quantum computer with user input.  

There are eight T-help subsets according to Ghosh [1] and Carbo et al [2]. The CD4+ cell may differentiate, into one of the eight subsets (T-help cells). It’ll be clear that when you read  “Pt” or “Pts”, I mean that the Protein [Fig-2][List of Abbreviations] is Tfh, Th9, Th2, iTreg, Tr1, Th22, Th17, or Th1. Also, a well-known differentiation is the evolution of a Naïve T-cell, to an effector T-cell or memory T-cell. Carbo’s team calls his [2] special differentiation process as “plasticity” if the CD4+ T-cell can evolve into one of the eight T- help Pts. This term is also a mathematical term. Eg., since Microsoft’s quantum computing  paradigm relies on tensors anyway – “for all observers in R³ eg, the basis vectors will transition from one reference frame to another in one a way. The components will change in such a way, that the combination of basis vectors and components will be the same for all observers.”   

                                                                      1

For each subset, I claim that the six attributes can be – attoamps [pg v][3], activation, expansion,  differentiation, expression, and plasticity. The coding is arbitrarily written for max_attoamps = 80 just for demonstration purposes. Also arbitrarily, min attributes = ”Expression”: 4, ”ACTivation”: 5, ”EXPansion”: 4, ”DIFFerentiation”: 5. “min_attributes” is one constraint of the application. Each attribute will have user input, and will contain values for each of the eight T-help proteins. Information will be explained in §4. The application is coded for goals being, ”ACTivation” and ”Plasticity”. ACTivation is ”weighed-against” Plasticity, (or vice-versa). Ie., deficit in one will be an asset for the other*. Three outcomes will be printed on the console output. The first one is for the best combination or “mix” of values for each Pt, given user input values, (assuming that there will be enough plasticity to begin with), there’s further information on this later still in §4.1. The second outcome is for the best ACTivation sample, and lastly, best Plasticity sample.

*The reason for this is because of the requirements of the Hamiltonian algorithm that’s used by D-WaveSys in their quantum annealing process [9].

Secondly – quantum computing is “a thing!”:

The Quantum Processing Unit (QPU) is accessed to perform the optimization calculations because of the complexity of the problem. It’s absolutely required if all 104 proteins are entered into the application’s input. For this work of course,  all eight (only) Carbo Pts are used for user input.

Please see Fig 3 for the Systems Biology Markup Language, (SBML) depiction of the CD4+ T-help cell compartment with its 104 related proteins (cytokines, chemokines, transcription factors, etc). Optimization results may never be found using conventional (AKA classical) compute that involves these 104 Pts.

Even if and when using HP computing architectures and multi-processes, the system will not give us the results in any reasonable amount of time. An example will be explained later regarding “number of operations” [pg 12].

I need to please make it clear that I am not a biologist nor a biology student. Moreover, I do not claim to be an adequate neophyte in biology, much less in the complex and  highly vastly nuanced field of immunology. Original inspiration for this work will be explained in the Related Work section, however this paper is a computer science paper, albeit a bit interdisciplinary. It is hoped however that the application that accompanies this work can be of some use to the immunology or cancer-research community, even if in the future. A citation will show my Github repository [6]. This repo should be cloned to include the SDK components, Ocean and dimod APIs. Contributions, questions, corrections, or improvements can be made or pushed to the repo.

https://github.com/TomEphraimPerez/learnleap

                                                                      2

Thank you : )